Spatiotemporal Change of Crop Yield and Its Response to Planting Structural Shifts in Northeast China from 2001 to 2021

Abstract

As a pivotal region for safeguarding China’s food security, Northeast China requires a quantitative evaluation of crop yield dynamics, planting structure shifts, and their interdependent mechanisms. Leveraging MODIS NPP data and remote sensing-derived crop classification data from 2001 to 2021, this study established a crop yield estimation model. By integrating the Theil–Sen median slope estimator and Mann–Kendall trend analysis, we systematically investigated the spatiotemporal characteristics of maize, rice, and soybean yields. Phased attribution analysis was further employed to quantify the effects of crop type conversions on total regional yield. The results revealed: (1) strong consistency between estimated yields and statistical yearbook data, with validation R² values of 0.76 (maize), 0.69 (rice), and 0.81 (soybean), confirming high model accuracy; (2) significant yield growth areas that spatially coincided with the core black soil zone, underscoring the productivity-enhancing role of conservation tillage practices; (3) all three crops exhibited upward yield trends, with annual growth rates of 1.33% (maize), 1.20% (rice), and 1.68% (soybean). Spatially, high-yield maize areas were concentrated in the southeast, rice productivity peaked along river basins, and soybean yields displayed a distinct north-high-south-low gradient; (4) crop type transitions contributed to a net yield increase of 35.9177 million tons, dominated by soybean-to-maize conversions (50.41% contribution), while maize-to-soybean shifts led to a 2.61% yield reduction. This study offers actionable insights for optimizing planting structures and tailoring grain production strategies in Northeast China, while providing a methodological framework for crop yield estimation in analogous regions.

1. Introduction

Food security remains a global priority, and accurate, real-time crop yield estimation is critical for agricultural policy formulation, resource allocation, and market stability. As the world’s largest producer and consumer of agricultural products, China bears the responsibility of ensuring food security for nearly 20% of the global population [1]. Northeast China, serving as the nation’s primary grain production base, accounts for over one-fifth of the national grain output and plays a critical role in safeguarding food security, with corn, rice, and soybean dominating its cropping systems. However, the spatiotemporal variability in crop growth and yield, driven by climate change and land use dynamics, has intensified the demand for reliable yield forecasting in this region. Optimizing cropping systems is essential for enhancing food security [2], reducing water consumption [3], and mitigating greenhouse gas emissions [4]. Quantifying the impacts of crop type shifts on regional productivity and systematically evaluating yield potential are thus imperative for ensuring sustainable agriculture and guiding evidence-based policymaking.

Crop yield estimation methods traditionally include statistical forecasting [5], agronomic surveys [6], numerical modeling [7], and remote sensing techniques [8]. Conventional approaches, such as administrative unit statistics and sampling surveys, provide reliable and detailed yield data but face limitations in scalability, cost-efficiency, and timeliness, particularly for large-scale regions. These constraints compromise the accuracy and practicality of yield predictions for real-world applications. Remote sensing-based methods, in contrast, offer advantages in macroscopic and real-time monitoring, enabling efficient and precise yield assessments over extensive areas [9].

The application of remote sensing technology, particularly in crop yield estimation utilizing MODIS NPP data, has achieved remarkable progress. For instance, it has been used to extract the features of winter wheat growth from multitemporal MODIS images for yield estimation in northern China [10]. A rice yield estimation model, developed based on MODIS NPP products and rice radiation efficiency, has been applied in Liling County [11], a major rice cultivation area in China; the utilization of extensive multivariate remote sensing data enabled precise county-level soybean yield estimation in the United States [12]; and annual summer crop planting area maps covering 14 provinces in Argentina were mapped using MODIS multi-temporal data combined with field-based crop reference records [13]. These methods leverage the continuous observational capabilities of remote sensing to track entire crop growth cycles, combining early growth status with remote sensing data enhances yield prediction accuracy, supporting preemptive yield assessments and disaster evaluations [14], thereby providing critical insights for policymakers. Although MODIS NPP and analogous datasets have been extensively utilized for crop monitoring, their coarse spatial resolution (e.g., 500 m) renders them insufficient in characterizing fine-scale cropland heterogeneity, while their capability to accurately estimate yields within complex agricultural systems remains limited. While significant progress has been made in agricultural research targeting Northeast China’s production systems, existing studies predominantly focus on either analyzing spatiotemporal variation characteristics of cropping patterns [15,16,17,18,19,20] or conducting single-crop yield estimations. There remains a notable gap in long-term continuous monitoring of multi-crop systems in this region, and particularly a scarcity of integrated studies combining annual crop yield estimation with cropping pattern dynamics.

Maize and soybean share overlapping growing seasons, similar climatic requirements, and substitutable roles in feed production [21]. However, low domestic soybean prices and profitability due to massive imports have driven farmers toward maize cultivation [22]. China’s soybean cultivation area decreased by over 400,000 hectares annually from 2006 to 2015 [23], leading to declining self-sufficiency. By 2017, 90% of China’s soybean demand relied on international imports [24]. In reality, soybean serves as a critical source of unsaturated fatty acids and proteins in human diets, stands as one of the most economically vital crops globally, and holds irreplaceable advantages over corn [25,26]. This planting structure imbalance has triggered structural contradictions in food security, exemplified by widespread soybean-to-maize conversions. A systematic evaluation of yield impacts from such crop type transitions is thus urgently needed.

This study aims to overcome the spatial limitations of traditional statistics by integrating remote sensing and ground observations to reconstruct spatiotemporally continuous yields of maize, rice, and soybean in Northeast China from 2001 to 2021. Validation using prefecture-level statistical yearbook data ensures model precision. By combining the Theil–Sen median slope estimator and Mann–Kendall trend test, we systematically quantify the spatiotemporal heterogeneity of yield changes. Addressing the research gaps in prior studies, this study uniquely evaluates multi-year continuous yields of maize, rice, and soybean in Northeast China while innovatively quantifying the contribution rate of cropping pattern shifts to total yield variations through phased attribution analysis. It further elucidates dynamic processes of productivity impacts induced by crop type transitions, thereby partially bridging critical knowledge gaps in agricultural system research. Our findings provide a quantitative basis for black soil conservation and cropping system optimization in Northeast China, while offering scientific and policy insights for national food security and agricultural sustainability.

2. Materials and Methods

2.1. Study Area

This study focuses on Northeast China, encompassing the provinces of Liaoning, Jilin, Heilongjiang, and the eastern parts of the Inner Mongolia Autonomous Region (including Chifeng, Tongliao, Xing’an League, and Hulunbuir) (Figure 1). Geographically, the region spans 38°43′ N to 53°33′ N latitude and 121°10′ E to 135°05′ E longitude, covering approximately 1.24 million km² [27]. As a vital commercial grain base in China, Northeast China is renowned for its production of maize, rice, and soybean. The region benefits from abundant agricultural resources and unique natural conditions that significantly influence crop growth.

The climate is predominantly temperate monsoon, transitioning from a mid-temperate zone in the south to a sub-frigid zone in the north. The mean annual temperature ranges from 4 to 7 °C (Figure 1a), while annual precipitation decreases from 750 mm in the southeast to 350 mm in the northwest (Figure 1b). These climatic characteristics profoundly shape crop growth cycles and yield patterns. Northeast China is characterized by fertile soils, particularly black soils (chernozems), which account for ~12% of the global black soil area. These soils are rich in humus and organic matter, providing optimal nutrient conditions for crops and underpinning the region’s status as a key grain-producing zone.

2.2. Data Sources

The data required for this study primarily included MODIS NPP data, annual crop type remote sensing data, crop yield statistics from statistical yearbooks, and administrative boundary data.

The MODIS NPP product is derived from remote sensing observations of photosynthetically active radiation (PAR), vegetation growth conditions, and land cover types, providing net primary productivity data. These products are provided by NASA and integrated into the Google Earth Engine (GEE) platform. Specifically, MOD16A2 represents monthly NPP data with a spatial resolution of 500 m (https://ladsweb.modaps.eosdis.nasa.gov/, accessed on 3 December 2024).

Crop type classification data were obtained from the Northeast China crop distribution dataset developed by Liu et al. [28]. This dataset integrates 111-dimensional MOD09A1 features on the GEE platform and employs feature optimization combined with a random forest classification algorithm to classify annual crops from 2000 to 2021. Validated against municipal and county-level statistical records, the dataset achieves an accuracy exceeding 85%, with a spatial resolution of 500 m. For this study, spatial distributions of maize, rice, and soybean were extracted (Figure 2).

Ground survey data, including measurements of drying coefficients, harvest indices, and root-to-shoot ratios, were derived from 111 sampling sites collected nationwide by Wang et al. during 2011–2012 [29]. For this study, geographically constrained selection was applied to retain survey data specific to Northeast China, ensuring regional specificity and applicability.

Crop yield statistics were collected from municipal statistical yearbooks across Northeast China (https://www.stats.gov.cn/sj/ndsj/, accessed on 9 December 2024), serving as reference data for validating yield estimation accuracy. Provincial and municipal administrative boundary data were acquired from the National Geomatics Center of China (https://www.ngcc.cn/, accessed on 7 December 2024).

2.3. Methods

2.3.1. Research Methods and Workflow

This study employed MODIS NPP data and a dry matter-to-yield conversion framework to conduct remote sensing estimation of crop yields and analyze the impact of crop type transitions on yield variations. The research consisted of two interconnected components: the first focuses on yield estimation for maize, rice, and soybean, while the second quantifies the contribution of crop planting pattern shifts to regional yield changes (Figure 3). The implementation process began with acquiring MODIS NPP datasets spanning 2001 to 2021 from the MODIS Earth Observation System, specifically targeting Northeast China’s agricultural growing seasons. Subsequent spatiotemporal matching of maize, rice, and soybean cultivation patterns was conducted according to their phenological characteristics, complemented by crop-specific parameters obtained through field surveys and synthesis of existing literature. These inputs enabled yield calculations using the dry matter conversion framework. To ensure robustness, long-term yield trends were systematically analyzed through the Theil–Sen median method and Mann–Kendall statistical test, with validation against authoritative agricultural yearbook records. Finally, a phased attribution analysis integrating historical crop planting structure data was implemented to disentangle and quantify the proportional impacts of crop type transitions on total yield dynamics. This integrated methodology enhances the mechanistic understanding of land use decisions in agricultural productivity, offering actionable insights for sustainable cropland management strategies.

2.3.2. Crop Yield Estimation Model

Net primary productivity (NPP) refers to the amount of organic matter accumulated by green plants per unit area and time, representing the carbon fixed through photosynthesis minus plant respiration losses. As a critical variable linking vegetation growth and crop yield, NPP is widely applied in vegetation monitoring, crop yield estimation, land use assessment, and ecological benefit evaluation [30]. Crop yield estimation models based on this principle are collectively termed NPP models, which are categorized into statistical models, process-based models, and parametric models. Among these, remote sensing-driven crop yield models that integrate dry matter–yield relationships have emerged as a research frontier due to their advantages in large-scale spatial coverage and dynamic monitoring capabilities. Consequently, this study adopts an NPP statistical model for crop yield estimation.

This study utilizes MODIS NPP data for large-scale crop yield estimation in the Northeast region, to some extent circumventing the issue of insufficient accuracy, and constructs a remote sensing model for crop yield estimation based on the crop dry matter–yield relationship [31]:

$$Y_i={NPP}_i\times {HI}_i\times p_i\times {\displaystyle \frac{a_i}{1-w_i}}$$

(1)

where $Y_i$ (g·m⁻²) denotes the estimated yield of crop i, ${NPP}_i$ represents the net primary productivity of crop i, ${HI}_i$ is the harvest index (proportion of harvestable biomass to total biomass), while $p_i$, $a_i$ and $w_i$ correspond to the carbon-to-biomass conversion coefficient, aboveground biomass ratio, and moisture content of crop i, respectively. This study optimizes the model by integrating prior research [29,32], field-measured data, and crop type distribution data to determine crop-specific parameters, thereby enhancing the reliability and accuracy of yield estimation outcomes.

2.3.3. Temporal Trend Detection Using Theil–Sen Median and Mann–Kendall Approaches

In crop yield estimation studies across Northeast China, the integrated application of the Theil–Sen median slope estimator and Mann–Kendall (MK) test aims to quantify long-term yield trends and assess the impacts of crop type changes. This combined approach is a widely recognized statistical method for detecting trend significance in time series data. The Theil–Sen median provides the magnitude of trends, quantifying annual change rates, while the MK test evaluates the statistical significance of these trends, ensuring robust conclusions.

The Theil–Sen median slope estimator calculates the annual rate of yield change per pixel or administrative unit from 2001 to 2021, expressed as a median slope (β). Results are used to compare productivity dynamics across regions and crop types.

The Mann–Kendall test identifies statistically significant increasing or decreasing trends. Regions with significant trends (p < 0.05) are classified as “improved” or “degraded” for targeted agricultural management.

Theil–Sen median formula [33]:

$$\beta =median\left({\displaystyle \frac{x_j-x_i}{j-i}}\right),\forall j>i$$

(2)

where $x_i$ and $x_j$ represent yield values at times (1 < i < j < n). A positive β indicates an upward trend, while a negative value denotes a downward trend. Notably, β reflects the median rate of change but does not assess statistical significance, necessitating complementary MK testing [34,35].

Statistic S:

$$S=\sum _{i=1}^{n-1}\sum _{j=i+1}^{n}sgn\left(x_j-x_i\right)$$

(3)

The sign function $sgn\left(x_j-x_i\right)$ is used to determine the relative changes in time series values, and its calculation formula is:

$$sgn\left(x_j-x_i\right)=\left\{\begin{array}{c}1 if(x_j-x_i>0)\\ 0 if(x_j-x_i=0)\\ -1 if(x_j-x_i<0)\end{array}\right.$$

(4)

Moreover, the statistic S follows a normal distribution, with a variance of:

$$Var\left(S\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-{\sum }_{i=1}^m{t}_i(t_i-1)({2t}_i+5)}{18}}$$

(5)

where m is the number of tied groups, and $t_i$ is the count of data points in the i-th group. When n ≥ 10, the calculation formula for Z is:

$$Z=\left\{\begin{array}{cc}{\displaystyle \frac{s-1}{\sqrt{Var\left(s\right)}}}& s>0\\ 0& s=0\\ {\displaystyle \frac{s+1}{\sqrt{Var\left(s\right)}}}& s<0\end{array}\right.$$

(6)

In this study, the time series spans 20 years, and the Z-statistic was employed for trend testing at a confidence level of α (α = 0.05). A trend was considered statistically significant if ∣Z∣ > 1.96.

2.3.4. Accuracy Validation

Accuracy validation of yield estimation results is essential to assess model reliability. In large-scale crop yield studies, statistical yearbook data are widely adopted as the benchmark for validation. To rigorously evaluate the consistency and accuracy between estimated yields and statistical records, this study employs four metrics: root mean square error (RMSE), normalized root mean square error (NRMSE), coefficient of determination (R²), and mean absolute percentage error (MAPE). Their formulas are defined as follows:

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}\sum _{i=1}^n{({Yield}_{estimated}\left(i\right)-{Yield}_{observed}\left(i\right))}^2}$$

(7)

where Yieldestimated(i) and Yieldobserved(i) represent the estimated and yearbook-reported yields for the i-th sample, respectively. Smaller RMSE values indicate higher estimation precision.

$$NRMSE={\displaystyle \frac{RMSE}{{Yield}_{observed}\left(max\right)-{Yield}_{observed}\left(min\right)}}$$

(8)

here, Yieldobserved(max) and Yieldobserved(min) denote the maximum and minimum observed yields, respectively. NRMSE standardizes RMSE to facilitate cross-region comparisons.

$$R^2={\displaystyle \frac{{\left({\sum }_{i=1}^n\left({Yield}_{estimated}(i\right)-\overline{{Yield}_{estimated}}\right)({Yield}_{observed}\left(i\right)-\overline{{Yield}_{observed}}))}^2}{{\sum }_{i=1}^n{\left({Yield}_{estimated}(i\right)-{Yield}_{estimated})}^2{\sum }_{i=1}^n{({Yield}_{observed}\left(i\right)-\overline{{Yield}_{observed}})}^2}}$$

(9)

where $\overline{{Yield}_{estimated}}$ and $\overline{{Yield}_{observed}}$ are the mean values of estimated and observed yields. An R² close to 1 signifies strong agreement, while values near 0 indicate poor consistency.

$$MAPE={\displaystyle \frac{1}{n}}\sum _{i=1}^n\mid {\displaystyle \frac{{Yield}_{estimated}\left(i\right)-{Yield}_{observed}\left(i\right)}{{Yield}_{estimated}\left(i\right)}}\mid$$

(10)

where n is the number of data points. MAPE quantifies the relative deviation between estimated and observed yields, with lower values reflecting higher accuracy.

3. Results

3.1. Accuracy Validation of Yield Estimation Results

Accuracy validation of yield estimates is a critical step to evaluate model reliability and applicability. To comprehensively assess the precision of maize, rice, soybean and total yield estimations across Northeast China, this study calculated four metrics (root mean square error—RMSE, normalized root mean square error—NRMSE, coefficient of determination—R², and mean absolute percentage error—MAPE) for the three crops over multiple years, generating robust accuracy assessments.

The validation results (

Table 1. Accuracy verification between estimated yield and observed yield.

Attribute	RMSE (kg·ha−1)	NRMSE (%)	MAPE (%)	R2
Maize	533.28	9.83	10.9	0.76
Rice	450.71	10.36	11.63	0.69
Soybean	284.65	9.74	9.64	0.81
Total	485.08	9.92	11.12	0.75

) demonstrated that maize yield estimations at the prefectural-city scale achieved an RMSE of 533.28 kg·ha⁻¹, NRMSE of 9.83%, MAPE of 10.9 and R² of 0.76 from 2001 to 2021, indicating relatively high accuracy and stability. For rice, the RMSE and R² were 450.71 kg·ha⁻¹ and 0.69, respectively, with NRMSE and MAPE values of 10.36 and 11.63%. While slightly less precise than maize, rice yield estimates remained sufficient for practical applications. Soybean yield estimations exhibited superior performance, with an RMSE of 284.65 kg·ha⁻¹, NRMSE of 9.74%, MAPE of 9.64 and R² of 0.81, outperforming both maize and rice in validation accuracy. For total crop yield, the RMSE and NRMSE were 485.08 kg·ha⁻¹ and 9.92%, respectively, with MAPE and R² values of 9.92 and 0.75. Overall, the estimated average yields showed strong consistency with statistical yearbook records, confirming high predictive reliability at the prefectural-city scale.

3.2. Temporal Analysis of Crop Yield Estimation Results

The Theil–Sen median slope estimator and Mann–Kendall analysis were employed to investigate the trends in average yields of maize, rice, and soybean across Northeast China from 2001 to 2021, with statistical significance tested at α = 0.05.

For maize yield variations (Figure 4a), increasing trends were widely distributed, with 31.3% of the regions exhibiting statistically significant growth (α = 0.05), primarily concentrated in the southeastern part of Northeast China. Slight increases were observed in 22.7% of the areas, mainly in the northern regions. In contrast, 10.5% of the regions showed significant yield declines, while 13.7% exhibited slight reductions, predominantly in the central and southwestern parts of Northeast China, particularly central Jilin Province and northern Liaoning Province. Approximately 23.8% of the regions maintained stable yield trends (Figure 4d).

Unlike maize, rice yield trends displayed more distinct spatial differentiation (Figure 4b). Areas with increasing yields were concentrated in the northern riverine zones, notably the Sanjiang Plain and Songnen Plain. Statistically significant increases (α = 0.05) were found in 30.1% of the study area, with an additional 18.0% showing slight growth. Conversely, declining trends were prominent in southern Northeast China, including southeastern Jilin Province and central/southeastern Liaoning Province, where 6.9% and 9.4% of areas experienced significant and slight decreases, respectively. Stable yield patterns accounted for 35.7% of the region (Figure 4d), concentrated in southern Jilin and central Liaoning Province.

Soybean yield changes exhibited a markedly different spatial pattern (Figure 4c). Significant variations were clustered in northern Northeast China, primarily within Heilongjiang Province, while the other three provinces largely maintained stable trends. Specifically, 11.7% of the regions in central/western Heilongjiang showed significant yield increases, and 5.1% exhibited slight growth in northwestern Heilongjiang. Significant and slight declines occupied 16.9% and 4.7% of the areas, respectively, distributed in central/northeastern Heilongjiang and the border zones between Jilin Province and Inner Mongolia. Notably, 61.6% of the regions displayed stable yield trends (Figure 4d), predominantly in central Northeast China, including southwestern Heilongjiang, central Jilin, and northwestern Liaoning.

Overall, the three crops demonstrated divergent yield dynamics: maize yields increased widely across the region, rice growth was concentrated in northern river basins, while soybean yields predominantly stabilized or declined, with limited growth restricted to central/western Heilongjiang.

3.3. Spatiotemporal Characteristics of Crop Yields

Figure 5a–c presents the annual average yields of individual crops. All three crops—maize, rice, and soybean—showed increasing trends. Linear fitting of multi-year average yields (Figure 5) indicated annual increments of 16.37 kg·ha⁻¹, 20.80 kg·ha⁻¹, and 22.03 kg·ha⁻¹ for maize, rice and soybean, respectively. Despite overall growth, interannual variability rates differed: maize, rice, and soybean exhibited rates of 1.33, 1.20, and 1.68% per year, respectively. Notably, crops with lower baseline yields demonstrated higher variability rates, reflecting a negative correlation between yield levels and interannual variability.

The annual average total yields of maize, rice, and soybean in Northeast China from 2001 to 2021 are illustrated in Figure 5d. The average overall yield per unit area exhibited an upward trend, with a linear fitting analysis revealing an annual increase of 59.20 kg·ha⁻¹ (1.32% per year). The average total yield rose from 7885.9 kg·ha⁻¹ in 2001 to 10,046.2 kg·ha⁻¹ in 2021, representing a cumulative growth of 2160.3 kg·ha⁻¹ (27.40%).

3.3.1. Spatiotemporal Differentiation of Average Overall Yield per Unit Area in Northeast China

The spatial distribution of total crop yields (Figure 6) revealed distinct gradients, with yields generally increasing from north to south and west to east. Maximum yields clustered in the southeastern and southern regions. From 2001 to 2011, minimum yields occurred near the provincial borders of Heilongjiang, Jilin, Liaoning, and Inner Mongolia, shifting to western Heilongjiang after 2013. The average total crop yields in Heilongjiang Province exhibited a westward-to-eastward increasing gradient, while Jilin Province demonstrated a northwest-to-southeast ascending trend. In Liaoning Province, yields radiated outward from central areas toward peripheral zones, and Inner Mongolia Autonomous Region displayed a distinct north-south dichotomy, with higher yields concentrated in its southern territories.

Crop-specific analyses revealed pronounced spatiotemporal disparities in the average yields of maize, rice, and soybean across Northeast China.

3.3.2. Spatiotemporal Differentiation of Average Maize Yield in Northeast China

Maize exhibited a consistent upward trend in average yield from 2001 to 2021 (Figure 7). The average maize yield increased by 27.31% (852.8 kg·ha⁻¹) from 3123 kg·ha⁻¹ in 2001 to 3975.8 kg·ha⁻¹ in 2021 (Figure 5a), accompanied by a 202.60% surge in total production from 36.134 to 109.3415 million tons. Spatially, yields generally increased from northwest to southeast, with high-yield clusters located in southeastern Jilin and eastern Liaoning. Low-yield zones predominated along the central Heilongjiang–Inner Mongolia border. Regionally, maize yields in Heilongjiang Province demonstrated a west-to-east increasing gradient, while Jilin Province exhibited a northwest-to-southeast ascending trend. Liaoning Province, with overall yield levels surpassing those of the other three provinces, displayed a centripetal expansion pattern from central areas to peripheral zones. The maize yields in the Inner Mongolia Autonomous Region exhibited an increasing trend from north to south. Overall, the regional disparities in maize yield are gradually narrowing across Northeast China.

3.3.3. Spatiotemporal Differentiation of Average Rice Yield in Northeast China

During 2001–2021, rice yields in Northeast China exhibited an overall increasing trend, with the average yield rising from 2963.3 kg·ha⁻¹ in 2001 to 3708.5 kg·ha⁻¹ in 2021, an increase of 745.2 kg·ha⁻¹ (25.15%) (Figure 5b). Total rice production surged by 97.30%, growing from 17.522 to 34.5711 million tons. Temporally, yield improvements progressed relatively slowly from 2001 to 2006 but accelerated markedly post-2006. Spatially, the low-yield zones of rice are predominantly concentrated in the border regions of Heilongjiang Province, Liaoning Province and the Inner Mongolia Autonomous Region. High-yield zones aligned predominantly with river systems, particularly within the three major plains (Sanjiang, Songnen and Liaohe), while yield gains expanded northward and eastward.

Regionally, Heilongjiang Province dominated this growth, exhibiting a southwest-to-northeast yield gradient, which accounted for the most substantial contribution to regional rice productivity changes. In Jilin Province, high-yield clusters concentrated in the central and northern riverine areas. The rice yields in Liaoning Province exhibited relatively minor intra-provincial disparities, with high-yield areas concentrated in the central Liaohe Plain and southeastern regions. In contrast, the Inner Mongolia Autonomous Region demonstrated limited rice cultivation, where spatial variation in yield remained indistinct (Figure 8).

3.3.4. Spatiotemporal Differentiation of Average Soybean Yield in Northeast China

Soybean yields in Northeast China exhibited an upward trajectory from 2001 to 2021, with average productivity increasing from 1799.6 to 2361.9 kg·ha⁻¹—a rise of 562.3 kg·ha⁻¹ (31.25%) (Figure 5c). Total production grew by 39.78%, climbing from 6.077 to 8.4943 million tons. Spatially, from 2001 to 2008, the soybean yields in Northeast China showed a distribution characteristic of being lower in the west and higher in the east, with the high-value areas located in the eastern part of Northeast China. After 2008, the soybean yields gradually shifted to a distribution pattern of higher yields in the north and lower yields in the south, and the regional differences in soybean yields gradually narrowed. The high-yield cores were concentrated in the north of Songnen Plain, contrasting with lower production in the southern part of the Inner Mongolia Autonomous Region and the northwest part of Jilin Province.

Regionally, Heilongjiang Province experienced the most pronounced variations. The soybean yields in Heilongjiang Province exhibited an overall upward trend from 2001 to 2005. During 2005–2013, a gradual downward trend emerged in soybean yields. Post-2013, the yield transitioned to a renewed upward trajectory. Jilin Province exhibited a northwest-to-southeast decreasing gradient, while Liaoning Province hosted localized high-yield clusters in its eastern sectors amid otherwise homogeneous yields. Inner Mongolia Autonomous Region maintained a north-high-south-low dichotomy, with yield maxima along its border with Heilongjiang (Figure 9).

3.4. Impacts of Crop Type Shifts on Production

3.4.1. Crop Distribution and Cultivation Area Dynamics

Crop distribution maps (Figure 2) reveal that maize cultivation expanded continuously from 2001 to 2021, becoming the dominant crop in Northeast China. Its planting areas were concentrated in the Liaohe and Songnen Plains, with a notable northward shift in cultivation boundaries. Rice cultivation also doubled in area during this period, primarily along riverine zones. Conversely, soybean cultivation contracted significantly, relocating to northern regions, particularly within Heilongjiang Province.

From the perspective of cultivation area conversions (Figure 10), maize cultivation expanded by approximately 14.30 million hectares (Mha) during 2001–2006, while soybean areas contracted by 13.00 Mha. This shift was driven by the conversion of ~13.88 Mha of cropland from soybean to maize, predominantly in the central Songnen Plain, including Qiqihar and Suihua cities.

From 2006–2011, soybean areas continued to decline but at a reduced rate, with 2.40 Mha converted to maize. Concurrently, maize areas grew modestly (+1.60 Mha), and rice cultivation expanded by 1.16 Mha, primarily through newly developed paddies in the northeastern Sanjiang Plain.

The period from 2011–2016 saw sustained growth in maize (+3.80 Mha, largely via soybean conversions) and rice (+1.56 Mha, concentrated in Heilongjiang Province), alongside a sharp contraction in soybean areas (−4.40 Mha).

During 2016–2021, maize further expanded (+3.20 Mha) into central Northeast China and the border zones between Inner Mongolia and Jilin/Heilongjiang provinces. Rice areas increased slightly (+0.61 Mha), while soybean cultivation declined by 0.92 Mha, with most converted land absorbed into maize production systems.

3.4.2. Impacts of Crop Type Conversions on Yield Dynamics

Crop type transitions in Northeast China during 2001–2021 exerted substantial impacts on total crop production. As detailed in

Table 2. Contribution of crop type conversions from 2001 to 2021 to the change in total crop production in Northeast China.

Conversion Type	2001–2021
	Change Amount (104t)	Contribution Rate (%)
Maize to Rice	−22.47	−0.39
Maize to Soybean	−150.34	−2.61
Maize to Others	−2.45	−0.04
Rice to Maize	152.88	2.66
Rice to Soybean	0.18	0.00
Rice to Others	−0.10	0.00
Soybean to Maize	2901.31	50.41
Soybean to Rice	−4.38	−0.08
Soybean to Others	−30.96	−0.54
Others to Maize	704.27	12.24
Others to Rice	12.71	0.22
Others to Soybean	31.12	0.54
Total	3591.77	62.40

, the cumulative yield variations induced by 12 major crop conversion pathways amounted to 35.9177 million tons, accounting for 62.40% of the region’s total yield increment over this period. Notably, soybean-to-maize conversions dominated these contributions, generating an additional 29.0131 million tons (50.41% of total gains). Conversely, maize-to-soybean transitions resulted in the largest yield reduction, decreasing total production by 1.5034 million tons (−2.61%).

As shown in Figure 11, the spatial distribution of the 12 crop type conversions across different phases reveals remarkable spatial differences. Overall, maize-oriented conversions are predominantly located in the central and southwestern parts of Northeast China. Rice-oriented conversions are mainly found in the northeastern and central riparian areas. Soybean-oriented conversions are concentrated at the junction of Jilin Province, Heilongjiang Province, and Inner Mongolia. By contrast, conversions to other crops are relatively limited, mostly appearing in southern Inner Mongolia and western Heilongjiang Province.

This study quantifies the impacts of 12 crop type conversion pathways on total crop production across four temporal phases, as summarized in

Table 3. Contribution of crop type conversions in four stages from 2001 to 2021 to the change in total crop output in Northeast China.

Type	2001–2006		2006–2011		2011–2016		2016–2021
	Change Amount (104t)	Contribution Rate (%)	Change Amount (104t)	Contribution Rate (%)	Change Amount (104t)	Contribution Rate (%)	Change Amount (104t)	Contribution Rate (%)
Maize to Rice	−57.49	−2.25	−200.25	−32.37	−160.83	−16.95	−202.03	−12.34
Maize to Soybean	−4.31	−0.17	−267.59	−43.26	−215.59	−22.72	−379.57	−23.18
Maize to Others	−8.19	−0.32	−41.19	−6.66	−75.11	−7.92	−30.18	−1.84
Rice to Maize	91.07	3.57	148.35	23.98	85.44	9.00	149.79	9.15
Rice to Soybean	1.51	0.06	1.56	0.25	−0.09	−0.01	−1.11	−0.07
Rice to Others	−0.18	−0.01	−0.71	−0.11	−0.37	−0.04	−0.77	−0.05
Soybean to Maize	1075.42	42.16	584.05	94.41	792.90	83.55	1127.43	68.85
Soybean to Rice	−72.04	−2.82	−14.04	−2.27	−1.81	−0.19	2.28	0.14
Soybean to Others	−106.35	−4.17	−41.71	−6.74	−66.00	−6.96	−10.06	−0.61
Others to Maize	199.16	7.81	106.05	17.14	83.08	8.76	261.08	15.94
Others to Rice	1.72	0.07	1.14	0.18	0.34	0.04	2.36	0.14
Others to Soybean	133.24	5.22	47.43	7.67	25.30	2.67	36.37	2.22
Total	1253.54	49.14	323.09	52.23	467.25	49.24	955.59	58.36

.

From 2001–2006, crop type transitions contributed a net yield increase of 12.5354 million tons (49.14% of the total production gain during this period). Soybean-to-maize conversions dominated these gains, generating an increase of 10.7542 million tons (42.16%). Conversely, soybean-to-other-crop transitions resulted in the largest reduction, decreasing total yields by 1.0635 million tons (−4.17%).

From 2006–2011, net yield gains from crop conversions reached 3.2309 million tons (52.23% of the period’s total increment). Soybean-to-maize transitions remained the primary driver, contributing 5.8405 million tons (94.41%). In contrast, maize-to-soybean conversions caused the most substantial decline (−2.6759 million tons, −43.26%), while maize-to-rice transitions reduced yields by 2.0025 million tons (−32.37%).

From 2011–2016, crop conversions added 4.6725 million tons to total yields (49.24% of the phase-specific growth). Mirroring prior trends, soybean-to-maize shifts accounted for 7.9290 million tons (83.55%) of gains, whereas maize-to-soybean reversions led to a reduction of 2.1559 million tons (−22.72%).

From 2016–2021, the final phase saw the largest net increase (9.5559 million tons, 58.36% of total growth), driven overwhelmingly by soybean-to-maize conversions (+11.2743 million tons, 68.85%). Maize-to-soybean transitions again caused the most pronounced losses (−3.7957 million tons, −23.18%).

4. Discussion

4.1. Accuracy and Applicability of the Yield Estimation Model

The crop yield estimation model developed in this study, based on MODIS NPP datasets and remote sensing-derived crop type data, demonstrated robust performance at the prefectural-city scale, with R² values of 0.76, 0.69, and 0.81 for maize, rice, and soybean, respectively. These results validate the reliability and utility of remote sensing data for large-scale crop yield estimation. Notably, soybean yield estimations exhibited superior accuracy compared to maize and rice, attributable to species-specific distribution patterns and parameter selection. Soybean cultivation in Northeast China is spatially concentrated, predominantly in northern Heilongjiang Province, and its relatively homogeneous canopy structure during growth stages minimizes mixed-pixel interference in remote sensing retrievals [36], enhancing the congruence of MODIS NPP data with field conditions. Furthermore, soybean’s light use efficiency and biomass conversion coefficients are less susceptible to environmental fluctuations, whereas maize and rice yields are more vulnerable to irrigation variability and extreme climatic events [37,38], thereby compromising estimation precision.

However, the 500-m spatial resolution of the MODIS NPP and crop type data may inadequately resolve small-scale intercropping or rotational fields, potentially leading to underestimated yields in such areas and introducing systemic biases into regional production assessments.

4.2. Drivers of Spatial–Temporal Changes in Crop Yields in the Northeast Region

From 2001 to 2021, the total maize yield in the Northeast region increased from 36.134 million tons to 109.342 million tons, a growth of 202.6%. Of this increase, 50.41% was due to the conversion from soybean to maize cultivation. The shift in planting types is the result of the combined effects of climate change and policy guidance. The average annual temperature in the Northeast region has risen by 0.3 °C in the past decade [39,40], pushing the northern boundary of maize cultivation northward by 200–300 km and expanding the suitable planting area. In the northern part of Heilongjiang Province, the newly added maize area accounted for about 65% of the total increase in the region. To boost farmers’ enthusiasm for maize cultivation and to mitigate the impact of the global financial crisis on domestic maize prices, the Chinese government implemented a temporary maize purchase and storage policy in the Northeast region in 2008. This policy enhanced farmers’ willingness to grow maize. Additionally, the market price of maize is generally 30–50% higher than that of soybeans, which also drove the shift from soybean to maize cultivation. Furthermore, with the promotion of maize varieties suitable for dense planting, maize yields have continued to increase, especially in high-yielding areas such as the southeastern part of Jilin Province. Mechanized farming and intensive fertilization have further unleashed the potential of maize production.

The expansion of rice cultivation areas and the increase in rice yields during this period were primarily driven by improvements in varietal adaptability and farmland infrastructure. The development and adoption of cold-tolerant rice varieties extended the northern limits of rice cultivation, with northeastern regions of Northeast China emerging as the most significant new rice growing zones. Concurrently, the construction and enhancement of field infrastructure, exemplified by projects in the Sanjiang Irrigation District, improved irrigation reliability and further unlocked rice production potential.

Despite a 31.25% increase in soybean yield per unit area, the substantial reduction in soybean acreage resulted in a much lower total output growth (39.78%) compared to maize and rice. This discrepancy can be attributed to two factors. First, prior to the implementation of the target price policy in 2014, low temporary procurement prices for soybean discouraged cultivation, prompting some farmers to shift to maize [41,42]. Second, China’s persistently high dependence on soybean imports [43] has destabilized domestic planting profitability, further hindering the recovery of soybean acreage.

4.3. Recommendations for Cropping Patterns

Based on quantitative analyses of spatiotemporal yield differentiation among major crops and the impacts of crop type transitions in Northeast China, this study reveals that the shift from soybean to maize and rice cultivation has driven regional yield increases. However, such imbalanced cropping patterns have exacerbated critical challenges, including black soil degradation, environmental pollution, and structural food security risks. Continuous maize monoculture has caused a persistent decline in organic matter content within the region’s black soils [44,45], with surface soil erosion already observed in certain areas. Excessive nitrogen fertilizer application in maize-dominated systems has led to substantial nitrogen losses via runoff, leaching, and volatilization, undermining the ecological sustainability of Northeast China’s black soil regions [46]. Furthermore, the expansion of rice cultivation, a major contributor to greenhouse gas emissions (accounting for approximately 48% of global crop-related emissions [47]), has intensified carbon pressures. Concurrently, the sustained reduction in soybean acreage has heightened China’s dependence on soybean imports, exposing domestic markets to international volatility and jeopardizing food security.

To address these issues, policymakers should prioritize the adoption of “maize–soybean rotation systems [48]”, supported by targeted subsidies for participating farmers, to stabilize soybean production, enhance self-sufficiency, and restore soil fertility. Conservation tillage practices, including no-till farming and straw incorporation, should be promoted to mitigate soil erosion and preserve soil structure. In rice growing areas, water-efficient techniques such as intermittent irrigation could reduce water consumption, while integrated agronomic strategies (e.g., optimized fertilizer use and methane reduction technologies) should be implemented to curb greenhouse gas emissions. These measures collectively aim to balance agricultural productivity with ecological resilience, ensuring long-term sustainability in Northeast China’s agroecosystems.

4.4. Limitations and Future Research

Although the crop yield estimation results in Northeast China effectively characterize the spatiotemporal patterns of corn, rice, and soybean production from 2001 to 2021, this study has several limitations that require further refinement in future work.

The crop yield estimation model constructed in this research employs fixed harvest indices and carbon biomass coefficients, failing to account for dynamic variations in crop types, genetic improvements in cultivars, and interannual climate fluctuations. Future studies should integrate dynamic parameters to enhance model accuracy [49]. The 500 m resolution of MODIS NPP data limits its capacity to detect variations in smallholder farming patterns, and incorporating higher resolution datasets like Sentinel-2 could improve estimation precision. Additionally, the analysis of crop distribution and yield changes in Northeast China lacks consideration of farmer decision-making behaviors (e.g., responses to subsidies) and specific climatic factors, which constrains in-depth exploration of driving mechanisms. Subsequent research should incorporate socioeconomic factors (e.g., agricultural policies and farmer decisions) to better elucidate yield variation drivers. The impacts of climate change on agricultural production constitute a global challenge. Given Northeast China’s status as a critical national grain production base, the responsiveness of its agricultural systems to climate shifts holds paramount significance, particularly as regional crop yields demonstrate heightened sensitivity to climate variability [50]. Future investigations should prioritize exploring the potential to systematically link agricultural practices in this region with broader climate change mitigation and environmental conservation frameworks.

5. Conclusions

This study developed a crop yield estimation model using MODIS NPP datasets and remote sensing crop classification data, generating annual maize, rice, and soybean yield datasets for Northeast China from 2001 to 2021. The interannual yield variations were analyzed using Theil–Sen median slope estimation and Mann–Kendall trend tests to characterize production trends for each crop. Furthermore, phased analyses were conducted to evaluate the impacts of crop type transitions on yield dynamics. The key findings are as follows:

(1)

Validation against prefecture-level statistical data demonstrated high accuracy of the yield estimation model, with R² values of 0.76, 0.69, and 0.81 for maize, rice, and soybean, respectively. These results highlight the reliability and practical utility of remote sensing data for crop yield estimation at regional scales.

(2)

The average yield changes of maize, rice, and soybean were classified into five categories. Maize yield growth primarily occurred in southeastern regions, with significantly increased areas accounting for 31.3%. For rice, 30.1% of regions showed significant yield increases, concentrated in northern riverine zones. Soybean yields remained stable in 61.6% of regions, while significant growth areas (11.7%) were predominantly located in north/central Heilongjiang Province.

(3)

From 2001 to 2021, the total average yield of maize, rice, and soybean in Northeast China increased from 7885.9 to 10,046.2 kg·ha⁻¹, representing a 27.39% growth. The annual change rates for these crops were 1.33, 1.20, and 1.68% per year, respectively. Maize yields increased by 852.8 kg·ha⁻¹ (27.31%), exhibiting a northwest-to-southeast increasing gradient. Rice yields rose by 745.2 kg·ha⁻¹ (25.12%), with high-yield areas aligned along river systems. Soybean yields demonstrated the most substantial growth at 562.3 kg·ha⁻¹ (31.25%), following a distinct north-high-south-low spatial pattern.

(4)

Crop type transitions contributed 62.40% to the total yield increment. Soybean-to-maize conversion emerged as the dominant driver, contributing 29.0131 million tons (50.41%) to yield gains. Conversely, maize-to-soybean transitions caused the largest yield reduction (−1.5034 million tons, −2.61%). Notably, soybean-to-maize shifts constituted the primary growth mechanism across the study period, except during 2001–2006 when maize-to-soybean conversions predominantly drove yield declines.